Former Baidu Chief Architect Crosses into Pharma: Wangshi Intelligence Leverages AI to Accelerate Drug Discovery

“10 years, $1 billion”: This is the classic summary of drug development. Drug development faces numerous pain points, with a development timeline of at least 10 years, an average cost of $2.63 billion, and a failure rate as high as 92%.

Against this backdrop, industry players have turned to AI to accelerate new drug development, thereby shortening development cycles and improving efficiency. Although the “AI + new drug development” trend has been prominent for four to five years, few competitors possess the capabilities to participate, leaving many under-the-radar projects yet to be uncovered.

Recently contacted by 36Kr"StoneWise", focusing on the AI-driven new drug development sector. The company was established in 2018 and had been engaged in technology development, remaining in an investment-intensive phase.The company is expected to publicly release its core technology platform—the Intelligent Drug Design Platform—for the first time in September this year.

This platform leverages artificial intelligence algorithms, such as generative adversarial networks (GANs) and autoencoders, integrated with computational chemistry knowledge, to intelligently design molecular structures. It comprehensively evaluates and ranks molecules based on their biological activity, DMPK properties, toxicity, and synthetic accessibility. Users can obtain high-quality molecule recommendations generated by the system through simple input operations. Currently, the team is constructing a proprietary patent knowledge graph using algorithms such as natural language processing (NLP) and optical structure recognition, ensuring that the recommended molecules do not infringe upon existing intellectual property rights.

StoneWise told 36Kr that the training data for its intelligent drug design platform comes from public platforms. In addition to the aforementioned capabilities, it essentially covers all stages from new drug discovery to preclinical research.

Beyond its intelligent platform, the company is also building a knowledge graph that covers drug patents, literature, and even first-hand data from within pharmaceutical companies. The knowledge graph is currently in its early stages. “We will adopt a strategy of advancing both the intelligent drug platform and the knowledge graph in tandem,” said CEO Zhou Jielong.

StoneWise has signed a collaboration agreement with the State Key Laboratory of Natural and Biomimetic Drugs’ Drug Design Research Division and, leveraging algorithmic design, has identified an active compound that is poised for subsequent experimental validation. Meanwhile, StoneWise and the MDL team from the Molecular Design Laboratory jointly won the championship in the finals of the “Merck Cup” Retrosynthesis Reaction Prediction Competition.

Compared with competitors, StoneWise’s advantage lies in its search and ranking capabilities. The team estimates that the chemical space of small-molecule compounds is on the order of 10^60. Identifying suitable molecules within this vast space is a highly challenging task, for which search and ranking are central. The key to new drug development resides in matching targets with molecules, leveraging computational methods to recommend optimal, druggable candidate molecules.

The company’s team possesses extensive expertise in search ranking. CEO Zhou Jielong previously served as a Principal Architect at Baidu and was one of the key figures driving technological innovation in Baidu Search., previously responsible for machine learning-based ranking in Baidu Search. During his tenure at Baidu, he led the team to rebuild the Baidu search engine using machine learning, continuously improving the long-tail query experience and enhancing search R&D efficiency. CEO Zhou Jielong stated that in 2013, his team successfully applied deep learning to search for the first time globally, two years ahead of Google.

Unlike some competitors that primarily focus on a First-in-Class strategy,StoneWise is currently focused on the research and development of Me-better drugs.This strategy was adopted because the technical framework for Me-better drugs is relatively more closed compared to First-in-Class drugs, making them easier to commercialize in the Chinese market. In the future, StoneWise will expand its pipeline to include Best-in-Class and First-in-Class candidates.

In the AI-driven new drug R&D industry, the core evaluation metric is the efficacy of drug discovery. What investors care about most is whether the drug candidates recommended by the platform can ultimately gain regulatory approval and reach the market. To address this, StoneWise is employing specific strategies for rapid validation, iterating a new version every one to two weeks on average.

As for future plans, StoneWise aims to deepen collaborations with several pharmaceutical companies through its platform in the second half of this year, and plans to expand into overseas pharmaceutical markets next year to provide First-in-Class drug development services.

StoneWise’s AI technical team comprises senior technical talent from leading internet and AI companies such as BAT, possessing core technological expertise in AI R&D fields—including natural language processing, computer vision, recommendation systems, search, and ranking systems—as well as in big data domains such as cloud storage and cloud computing. Its drug discovery team consists of PhDs and Masters specializing in computational biology, cheminformatics, synthetic chemistry, pharmacy, and medicine, ensuring professional expertise across the entire drug development lifecycle. The company’s advisory board includes professors of pharmacy from Peking University, former medicinal chemists from Merck, and computational chemists.

StoneWise completed a multi-million US dollar angel financing round at its inception, with funding from SIG (Susquehanna International Group).Currently, the company is undertaking a new round of financing, with the funds to be used for establishing overseas offices, expanding its team, broadening market reach, and enriching its database resources.